Comment Moderation: My Thoughts
Scott posted looking for feedback on comment moderation ideas. This is cheating on my part since I had this discussion with Scott a few months ago, but I'll repost my thoughts about comment control here because its a horizontal notion that applies to other applications in this space.
Simple single-state, single-actor moderation is fine. That should be included as a baseline for site owners who want a simple process that gives them full control and that everyone can understand. So that's a requirement.
Additionally, I'd extend this by adding email notificaton for comments whereby the site owner can remove comments with a reply back to an alert email--sort of a passive moderation mode that allows comments to flow-through but reduces the level of effort/mode-shifting required by the owner.
Given that the next version of .Text/CSB is going to support multiple authors, adding single-state, multiple-actor moderation should be fairly straightforward. I think this is a nice second level of moderation which will allow high-volume sites to have straightforward moderation that is less gating in terms of conversation flow.
Lastly, I think an additional moderation model resembling that used by /. and other established community sites should be adopted. While sites like /. and kuro5hin have more news-oriented roots, the same conceptual model applies. Functional components:
- Users can register as known users with a site or community (or theoretically, down the road, within a federated community, similar to what Six Apart are attempting).
- All posts and comments are rated on a fixed scale. For purposes of illustration we'll go with a 0-5 scale.
- New posts from unknown authors start with a rating of 1.
- Post and comment listings are equipped with a view filter. Users can choose what minimum rating threshhold will be required for a post or comment to appear on their listing. The filter's default minimum rating would be 1 in our example. Filter values are persisted for users (say, cookies) within sites/communities.
- Registered users can rate comments and posts, from 0 to 5.
- A minimum number of user ratings is required before a post or comment's rating is official. Let's say 3 ratings are required for this example. Until the minimum number of user ratings is reached, the initial rating is used for filtering (in this case, we're using 1).
Non-spam comment is made: Until three registered users decide it's worthless and rate it 0, it appears in comment listings (by default).
Spam comment is made: Same story. Once three registered (and this number can slide down on lower traffic sites I suppose), users give it a 0 it's hidden from default view. As an extension, this could trigger a notification event to the admin to passively or actively approve deletion.
In the meantime, users can filter out low quality posts and or comments by setting their default filter to 2+.
This won't do much about comment spam in practical terms, but we have made comment spam self moderating. But where it starts to make a ton of sense is in a large/active community. Take weblogs.asp.net. I would love to read it regularly but squelch the noise way down. I don't need to see the same 18 posts about the latest thing released on MSDN, pictures of someone's weekend or the many other types of repetitive, less informative posts--I swear half the time people are operating on some sort of personal quota with posts or just caught up in some name dropping dynamic. But I digress. There's a lot of good info being generated, we need better tools to cut down on the background chatter and zero in the useful stuff. Ratings are one way to do that.
If you filtered out only say 4 or 5 star items on weblogs.asp.net, I'm sure you're looking at only the primo stuff, the really good posts or the first instance of a post that gets echoed heavily. The next item on the agenda would be to translate the UX paradigm into filtered feeds. I would subscribe to a weblogs.asp.net 5-star feed, without some noise filter, there's no point.
Addtional: CAPTCHA support is easy to implement and fairly effective in blocking a lot of the comment posting agents people use. This should be added as an optional wrinkle for membership registration and for comment posting on sites configured to allow anonymous comments.
Anyway, that's my random babbling about the matter. For my part, I think we're moving to a world where only known community members can contribute. Which will increase transactional impedence for community interaction somewhat. At the same time, what is sum of the real concrete value coming from anonymous comments--I'd say it's pretty minimal to begin with and if registration were the norm already, this would be a largely moot issue. The real focus is how to layer a single sign-on mechanism across sites and engines... hey, I know, why not Passport!? :)